SlideShare a Scribd company logo
Embedded CDNs in 2023
﹣A status
Jack Fang
June 2023
What is a CDN?
Embedded servers are when some of the
CDNs servers are located within ISPs networks
and intended to only serve end-users in that
network. The servers will use address space
originated in the ISPs network and not in the
CDNs ASN(s).
Clustering is when a number of servers share
their storage. This means that the content is
distributed over the servers in the cluster and
will not be present at all of them. Sometimes
servers in different sites can be clustered
together.
A collections of
geographically
distributed servers
A method to place
content on the servers
A method to steer end-
users to the closest
servers
A bit about the CDN market
Growth in the market based on
the increased consumer
demand for entertainment and
OTT services and in the later
years on increased demand for
e-learning and work from
home solutions.
The traditional players like
Akamai, Lumen, Tata, Edgio
etc are well established as
reliable and with global reach.
The newer major CDNs like
Cloudflare, Fastly and
Stackpath are offering a more
diverse and specialised set of
services like security services
and DDOS protection, high
performance real time delivery
services etc.
Which btw is also offered by
Akamai
Large content producers who
use comercial CDNs often use
multiple CDNs. This both to
support their different types of
content, to leverage different
geographical scope of the CDN,
create resilience and finally -
more than one provider means
better leverage when
negotiating contracts…
A bit about the CDN market
The cloud providers allmost all
have a CDN to front their
content
The CDNs are almost all also
offering compute and other
cloud services
So in the end they end up
offering the same services but
with different specialities and
strengths
The giant content producers
like Netflix, Apple, Microsoft,
Amazon, Facebook and Google
have all build their own
specialised CDN to support
their core services ( and some
also sell CDN services…)
Building a CDNs on a global
scale is not small task, so this
solution is only for the largest
and most resourceful content
producers with enough
demand for traffic to make the
business case work.
Who offers embedded servers?
● Akamai
● Netflix
● Google
● Amazon
● Facebook
● Cloudflare
● CDN77
● Microsoft
● Apple
● Qwilt
● And more to
come
So, what is in it for the ISP?
Statistics
• Petros Gigis, Lefteris Mannasakis
and more presented at
Sigcomm’21 that over 3,500
ASNs globally have embedded
servers
• Most ISPs in the Kentik traffic
sample have 6 or less embedded
CDNs
• The top 6 in the Kentik sample
are:
Percentages of the ISPs in the sample that
have embedded servers from the CDN
Google (25.5%)
Facebook (16.2%)
Netflix (26.5%)
Akamai (11.6%)
Apple (3.3%) CDN77 (3.3%)
Total distribution of traffic
marked as CDN traffic
~74% over the external
network border
~14% from embedded servers
~7% embedded to embedded
internally in the networks
~5% Externally to
embedded
Average traffic over 24 hours that includes a sunday evening.
Statistics
● Outside to end-users (39%)
● Embedded to end-users (56%)
● Outside to embedded (5%)
● Embedded to embedded (1%)
Google Netflix Facebook Akamai
● Outside to end-users (22%)
● Embedded to end-users (72%)
● Outside to embedded (3%)
● Embedded to embedded (3%)
● Outside to end-users (43%)
● Embedded to end-users (54%)
● Outside to embedded (4%)
● Embedded to embedded (0%)
● Outside to end-users (73%)
● Embedded to end-users (18%)
● Outside to embedded (9%)
● Embedded to embedded (0%)
Steer to the closest cache
● BGP is used to signal the subnets
that a cache or cluster should
serve
● The most control is given
for the systems where every
cache has a BGP session and
the steering system respects
common BGP parameters.
● A setup where all Servers have
an identical session will work
for most deployments.
● Prefixes are sent from the ISP
to the cache
BGP
● End-users are mapped to a
cluster based on the DNS
server they are using to
request the content
DNS Server
● Geolocation is used by some
as the primary way if mapping an
end-user to a cluster. Some use
the geolocation of the end-user IP
address, but some rely on the
geo location of the DNS server the
end-user is using to request the
content.
Geolocation
Anycast
● Anycast addresses are
announced from the
caches and the ISPs
routing decides which is
the closest
BGP
x.y.z.0/24
x.y.w.0/24
x.y.w.123
B
1 2
3
4
A
● x.y.z.0/24 is announced to A
● x.y.w.0/24 is announced to B
1. Give me movie
2. Go to B and get movie
3. Give me movie
4. Movie
This works really well for ISPs who have
a very strict regional address plan. It
does not work for ISPs where they do not
have control over IP address
assignments or who do not have regional
aggregation
1. Where is site.com?
2. Where is site.com?
3. site.com is
site.com.cdnsomething.com
4. Where is site.com.cdnsomething.com
5. site.com.cdnsomething.com is A for
you
6. Site.com is A
7. Give me Site.com
8. Site.com
DNS
User in
Mumbai
Local
DNS
Authorized
DNS
CDN
node in Mumbai
IP address “A”
CDN CDN
node in Singapore
IP address “B”
2
3
1
6
7 8 4 5
And then some of the magic
● The CDN creates a mapping
of ISP DNS servers to POPs or
clusters based on latency
measurements
● Some CDNs include internet
connectivity from the cluster
to the DNS servers or IP
addresses in their mapping
algorithm
● The content owned CDNs
also include signals from the
content clients in the
algorithm
Latency and other QoS data
● The load of the
individual servers or
clusters is part of the
decision of exactly
which server the end-
user will end up using.
● Load balancing can take
place in the cluster or in
the CDN response to
(where is?)
Load
● Not all content is on all
Embedded servers
● Content that is pushed out is
distributed according to CDN
magic
● Not all content from a CDN will be
cached when the content is
pulled
● This means where the content
is/can be is also part of the
steering decision.
Content
Traffic profiles for
different ways of
placing content
• Centrally calculated
placement
• Fill window
• Optimized to reduce
edge fill traffic as much
as possible
Typical Netflix traffic profile in well dimensioned deployment
Traffic profiles for
different ways of
placing content
• Proxy and akamai magic
• Relatively dynamic
content
• Not all content will
be at the embedded
servers
Typical Akamai traffic profile in well dimensioned deployment
Traffic profiles for different ways of placing content
• Proxy and Facebook magic
• Relatively dynamic content
• Only the static is cached
• Not all content will ever
be at the embedded servers
Typical Facebook traffic profile in well dimensioned deployment
Feeds
Likes
Anything that needs to
be globally consistent
Limit who can use the Servers
Most embedded Servers will
have access list that limits
who will be served by the
cache
The common source to
generate these
is a BGP feed from the host
ISP
Some will use one single feed
per ISP, other per cluster or
per cache.
There must be consistency between the user-
group that can be mapped to a cluster or server
and the user-group that are allowed to use the
cluster or server.
This means there is a requirement for the CDNs
where each server has a BGP sessions that the
announcements to each server in a cluster is
identical.
Deploying embedded servers deep into the network
• Understand if the end-user
mapping is supported by your
network implementation
• DNS servers
• IP address plan
• Use of CGN
• Be aware of clustering
• Off load/hit rate for the local
cluster might be lower than for a
big cluster because of less disk
space in the cluster
And what about the hardware?
1-2U
10G /50G/100G Interfaces
Choice of AC and DC power
It varies whether a CDN managed switch is
required optional or not applicable.
Installation process are more/or less
automated, but some still require the ISP
tech to do basic configuration of the CDN
equipment.
Evolves continuously: less space,
more throughput, more and faster
storage Interfaces supports the
industry standard at any given time
Is it easier or harder in 2023 compared to 2012?
● Large boxes
● Off load unpredictable
● Few players
● No CGN support
2012
● Small boxes
● More predictable offload
● More predictable and flexible
end-user mapping
● Many players
● Some CGN support
2023
CONCLUSION
What do you experience?
THANK YOU
Jack Fang
Solutions Engineer - APAC
jfang@kentik.com

More Related Content

PDF
Edge virtualisation for Carrier Networks
PDF
Cloud SDN: BGP Peering and RPKI
PDF
Introducing Peering LAN 2.0 at DE-CIX
PDF
Peering Personal MyNOG-10
PDF
Hierarchical Network Controller
PDF
SDN & NFV Introduction - Open Source Data Center Networking
PDF
Diameter Overview
PPTX
Introduction to nexux from zero to Hero
Edge virtualisation for Carrier Networks
Cloud SDN: BGP Peering and RPKI
Introducing Peering LAN 2.0 at DE-CIX
Peering Personal MyNOG-10
Hierarchical Network Controller
SDN & NFV Introduction - Open Source Data Center Networking
Diameter Overview
Introduction to nexux from zero to Hero

What's hot (20)

PPT
VPN presentation - moeshesh
PDF
WiFi – Mobile BNG Offload Deployments
PDF
Understanding Cisco’ Next Generation SD-WAN Technology
PDF
The business case for SD WAN in the enterprise
PDF
FUTURE-PROOFING DATA CENTRES from Connectivity Perspective
PPTX
Cloud platform technical sales presentation
PDF
Malaysia Data Center Landscape, Where is the next hotspot to place your fiber...
PDF
SDM – A New (Subsea) Cable Paradigm
PDF
Easy, Secure, and Fast: Using NATS.io for Streams and Services
PPTX
cloud-migrations.pptx
PDF
Cisco umbrella overview
DOCX
Packet Tracer: Load Balancing with GLBP and FHRP
PDF
Aether: The First Open Source 5G/LTE Connected Edge Cloud Platform
PDF
Enterprise WAN Transformation: SD-WAN, SASE, and the Pandemic
PDF
TechWiseTV Workshop: Cisco SD-WAN
PDF
Network Virtualization in Cloud Data Centers
PPTX
SD-WAN 2.0: Building a Better SD-WAN
PPTX
Vxlan control plane and routing
PDF
Container Networking Deep Dive
PDF
SDWAN vs MPLS: What Enterprises need?
VPN presentation - moeshesh
WiFi – Mobile BNG Offload Deployments
Understanding Cisco’ Next Generation SD-WAN Technology
The business case for SD WAN in the enterprise
FUTURE-PROOFING DATA CENTRES from Connectivity Perspective
Cloud platform technical sales presentation
Malaysia Data Center Landscape, Where is the next hotspot to place your fiber...
SDM – A New (Subsea) Cable Paradigm
Easy, Secure, and Fast: Using NATS.io for Streams and Services
cloud-migrations.pptx
Cisco umbrella overview
Packet Tracer: Load Balancing with GLBP and FHRP
Aether: The First Open Source 5G/LTE Connected Edge Cloud Platform
Enterprise WAN Transformation: SD-WAN, SASE, and the Pandemic
TechWiseTV Workshop: Cisco SD-WAN
Network Virtualization in Cloud Data Centers
SD-WAN 2.0: Building a Better SD-WAN
Vxlan control plane and routing
Container Networking Deep Dive
SDWAN vs MPLS: What Enterprises need?
Ad

Similar to Embedded CDNs in 2023 (20)

PDF
PLNOG 6: Christian Kaufmann - How Akamai delivers your packets - the insight
PDF
PLNOG 6: Christian Kaufmann - How Akamai delivers your packets - the insight
PPTX
History of CDN - GlobalDots
PPT
PDF
Traffic Engineering for CDNs
PDF
InfoWorld
PDF
130815 - Content Delviery Networks for the IEEE Singapore Broadcast group
PDF
Content Delivery Network - Exploring the Power
PDF
Traffic Engineering for CDNs
PPTX
crit_think_intro classroom slides class notes
PDF
Cdn prospects 2011
PPTX
Traffic Engineering for CDNs by Matt Jansen [APRICOT 2015]
PDF
Akamai company profile
PPT
CONTENT DELIVERY NETWORK
PPTX
Effective use of CDNs
PPTX
Contents Localization Initiatives to get better User Experience
PPTX
Content Devilery Network
PDF
SPA2015: Hooman Beheshti – The Future of CDNs
PLNOG 6: Christian Kaufmann - How Akamai delivers your packets - the insight
PLNOG 6: Christian Kaufmann - How Akamai delivers your packets - the insight
History of CDN - GlobalDots
Traffic Engineering for CDNs
InfoWorld
130815 - Content Delviery Networks for the IEEE Singapore Broadcast group
Content Delivery Network - Exploring the Power
Traffic Engineering for CDNs
crit_think_intro classroom slides class notes
Cdn prospects 2011
Traffic Engineering for CDNs by Matt Jansen [APRICOT 2015]
Akamai company profile
CONTENT DELIVERY NETWORK
Effective use of CDNs
Contents Localization Initiatives to get better User Experience
Content Devilery Network
SPA2015: Hooman Beheshti – The Future of CDNs
Ad

More from MyNOG (20)

PDF
MEASURING THE HEALTH AND RESILIENCE OF THE INTERNET: MALAYSIA
PDF
Malaysia’s Emerging Trends in Data Center: Identifying Tomorrow’s Hotspots
PDF
SHADOWSERVER: INTERNET CRITICAL SECURITY AS A PUBLIC SERVICE
PDF
Building a Connected Future: The Power of Interconnection
PDF
COHERENT OPTICAL TRANSCEIVERS – CURRENT CAPABILITIES AND FUTURE POSSIBILITIES
PDF
Strategies for Seamless Recovery in a Dynamic Data Landscape
PDF
SRv6: DEPLOYMENT & USECASES by Aditya Kaul
PDF
Equinix: New Markets, New Frontiers
PDF
Securing the Onion: 5G Cloud Native Infrastructure
PDF
Cleaning up your RPKI invalids
PDF
Load balancing and Service in Kubernetes
PDF
AI in Networking: Transforming Network Operations with Juniper Mist AIDE
PDF
Keep Ukraine Connected: A project from the community – for the community by R...
PDF
Solving Civilization’s Long Term Communication Needs by Dinesh Kummaran, Tran...
PDF
MyIX Updates by Raja Mohan Marappan, MyIX
PDF
Exploring Quantum Engineering for Networking by Melchior Aelmans, Juniper Net...
PDF
Quick wins in the NetOps Journey by Vincent Boon, Opengear
PDF
Data Centre Interconnect (DCI) with X86’s DCI Solution by Raja Akmal, X86 Net...
PDF
Routed Optical Networking by Shahnaz Mohamad, Cisco
PDF
Edge Computing: NTT Offerings in Japan and Use Cases by Katsuhiro Ohki, NTT L...
MEASURING THE HEALTH AND RESILIENCE OF THE INTERNET: MALAYSIA
Malaysia’s Emerging Trends in Data Center: Identifying Tomorrow’s Hotspots
SHADOWSERVER: INTERNET CRITICAL SECURITY AS A PUBLIC SERVICE
Building a Connected Future: The Power of Interconnection
COHERENT OPTICAL TRANSCEIVERS – CURRENT CAPABILITIES AND FUTURE POSSIBILITIES
Strategies for Seamless Recovery in a Dynamic Data Landscape
SRv6: DEPLOYMENT & USECASES by Aditya Kaul
Equinix: New Markets, New Frontiers
Securing the Onion: 5G Cloud Native Infrastructure
Cleaning up your RPKI invalids
Load balancing and Service in Kubernetes
AI in Networking: Transforming Network Operations with Juniper Mist AIDE
Keep Ukraine Connected: A project from the community – for the community by R...
Solving Civilization’s Long Term Communication Needs by Dinesh Kummaran, Tran...
MyIX Updates by Raja Mohan Marappan, MyIX
Exploring Quantum Engineering for Networking by Melchior Aelmans, Juniper Net...
Quick wins in the NetOps Journey by Vincent Boon, Opengear
Data Centre Interconnect (DCI) with X86’s DCI Solution by Raja Akmal, X86 Net...
Routed Optical Networking by Shahnaz Mohamad, Cisco
Edge Computing: NTT Offerings in Japan and Use Cases by Katsuhiro Ohki, NTT L...

Recently uploaded (20)

PDF
Architecting across the Boundaries of two Complex Domains - Healthcare & Tech...
PDF
Peak of Data & AI Encore- AI for Metadata and Smarter Workflows
PPTX
Effective Security Operations Center (SOC) A Modern, Strategic, and Threat-In...
PDF
NewMind AI Monthly Chronicles - July 2025
PDF
Machine learning based COVID-19 study performance prediction
PDF
7 ChatGPT Prompts to Help You Define Your Ideal Customer Profile.pdf
PPTX
KOM of Painting work and Equipment Insulation REV00 update 25-dec.pptx
PDF
How UI/UX Design Impacts User Retention in Mobile Apps.pdf
PDF
NewMind AI Weekly Chronicles - August'25 Week I
PPTX
Big Data Technologies - Introduction.pptx
PDF
Modernizing your data center with Dell and AMD
PPTX
Cloud computing and distributed systems.
PDF
Blue Purple Modern Animated Computer Science Presentation.pdf.pdf
PPTX
PA Analog/Digital System: The Backbone of Modern Surveillance and Communication
PDF
KodekX | Application Modernization Development
PDF
CIFDAQ's Market Insight: SEC Turns Pro Crypto
PDF
Agricultural_Statistics_at_a_Glance_2022_0.pdf
PDF
Building Integrated photovoltaic BIPV_UPV.pdf
PDF
Mobile App Security Testing_ A Comprehensive Guide.pdf
PPTX
VMware vSphere Foundation How to Sell Presentation-Ver1.4-2-14-2024.pptx
Architecting across the Boundaries of two Complex Domains - Healthcare & Tech...
Peak of Data & AI Encore- AI for Metadata and Smarter Workflows
Effective Security Operations Center (SOC) A Modern, Strategic, and Threat-In...
NewMind AI Monthly Chronicles - July 2025
Machine learning based COVID-19 study performance prediction
7 ChatGPT Prompts to Help You Define Your Ideal Customer Profile.pdf
KOM of Painting work and Equipment Insulation REV00 update 25-dec.pptx
How UI/UX Design Impacts User Retention in Mobile Apps.pdf
NewMind AI Weekly Chronicles - August'25 Week I
Big Data Technologies - Introduction.pptx
Modernizing your data center with Dell and AMD
Cloud computing and distributed systems.
Blue Purple Modern Animated Computer Science Presentation.pdf.pdf
PA Analog/Digital System: The Backbone of Modern Surveillance and Communication
KodekX | Application Modernization Development
CIFDAQ's Market Insight: SEC Turns Pro Crypto
Agricultural_Statistics_at_a_Glance_2022_0.pdf
Building Integrated photovoltaic BIPV_UPV.pdf
Mobile App Security Testing_ A Comprehensive Guide.pdf
VMware vSphere Foundation How to Sell Presentation-Ver1.4-2-14-2024.pptx

Embedded CDNs in 2023

  • 1. Embedded CDNs in 2023 ﹣A status Jack Fang June 2023
  • 2. What is a CDN? Embedded servers are when some of the CDNs servers are located within ISPs networks and intended to only serve end-users in that network. The servers will use address space originated in the ISPs network and not in the CDNs ASN(s). Clustering is when a number of servers share their storage. This means that the content is distributed over the servers in the cluster and will not be present at all of them. Sometimes servers in different sites can be clustered together. A collections of geographically distributed servers A method to place content on the servers A method to steer end- users to the closest servers
  • 3. A bit about the CDN market Growth in the market based on the increased consumer demand for entertainment and OTT services and in the later years on increased demand for e-learning and work from home solutions. The traditional players like Akamai, Lumen, Tata, Edgio etc are well established as reliable and with global reach. The newer major CDNs like Cloudflare, Fastly and Stackpath are offering a more diverse and specialised set of services like security services and DDOS protection, high performance real time delivery services etc. Which btw is also offered by Akamai Large content producers who use comercial CDNs often use multiple CDNs. This both to support their different types of content, to leverage different geographical scope of the CDN, create resilience and finally - more than one provider means better leverage when negotiating contracts…
  • 4. A bit about the CDN market The cloud providers allmost all have a CDN to front their content The CDNs are almost all also offering compute and other cloud services So in the end they end up offering the same services but with different specialities and strengths The giant content producers like Netflix, Apple, Microsoft, Amazon, Facebook and Google have all build their own specialised CDN to support their core services ( and some also sell CDN services…) Building a CDNs on a global scale is not small task, so this solution is only for the largest and most resourceful content producers with enough demand for traffic to make the business case work.
  • 5. Who offers embedded servers? ● Akamai ● Netflix ● Google ● Amazon ● Facebook ● Cloudflare ● CDN77 ● Microsoft ● Apple ● Qwilt ● And more to come
  • 6. So, what is in it for the ISP?
  • 7. Statistics • Petros Gigis, Lefteris Mannasakis and more presented at Sigcomm’21 that over 3,500 ASNs globally have embedded servers • Most ISPs in the Kentik traffic sample have 6 or less embedded CDNs • The top 6 in the Kentik sample are: Percentages of the ISPs in the sample that have embedded servers from the CDN Google (25.5%) Facebook (16.2%) Netflix (26.5%) Akamai (11.6%) Apple (3.3%) CDN77 (3.3%)
  • 8. Total distribution of traffic marked as CDN traffic ~74% over the external network border ~14% from embedded servers ~7% embedded to embedded internally in the networks ~5% Externally to embedded Average traffic over 24 hours that includes a sunday evening.
  • 9. Statistics ● Outside to end-users (39%) ● Embedded to end-users (56%) ● Outside to embedded (5%) ● Embedded to embedded (1%) Google Netflix Facebook Akamai ● Outside to end-users (22%) ● Embedded to end-users (72%) ● Outside to embedded (3%) ● Embedded to embedded (3%) ● Outside to end-users (43%) ● Embedded to end-users (54%) ● Outside to embedded (4%) ● Embedded to embedded (0%) ● Outside to end-users (73%) ● Embedded to end-users (18%) ● Outside to embedded (9%) ● Embedded to embedded (0%)
  • 10. Steer to the closest cache ● BGP is used to signal the subnets that a cache or cluster should serve ● The most control is given for the systems where every cache has a BGP session and the steering system respects common BGP parameters. ● A setup where all Servers have an identical session will work for most deployments. ● Prefixes are sent from the ISP to the cache BGP ● End-users are mapped to a cluster based on the DNS server they are using to request the content DNS Server ● Geolocation is used by some as the primary way if mapping an end-user to a cluster. Some use the geolocation of the end-user IP address, but some rely on the geo location of the DNS server the end-user is using to request the content. Geolocation Anycast ● Anycast addresses are announced from the caches and the ISPs routing decides which is the closest
  • 11. BGP x.y.z.0/24 x.y.w.0/24 x.y.w.123 B 1 2 3 4 A ● x.y.z.0/24 is announced to A ● x.y.w.0/24 is announced to B 1. Give me movie 2. Go to B and get movie 3. Give me movie 4. Movie This works really well for ISPs who have a very strict regional address plan. It does not work for ISPs where they do not have control over IP address assignments or who do not have regional aggregation
  • 12. 1. Where is site.com? 2. Where is site.com? 3. site.com is site.com.cdnsomething.com 4. Where is site.com.cdnsomething.com 5. site.com.cdnsomething.com is A for you 6. Site.com is A 7. Give me Site.com 8. Site.com DNS User in Mumbai Local DNS Authorized DNS CDN node in Mumbai IP address “A” CDN CDN node in Singapore IP address “B” 2 3 1 6 7 8 4 5
  • 13. And then some of the magic ● The CDN creates a mapping of ISP DNS servers to POPs or clusters based on latency measurements ● Some CDNs include internet connectivity from the cluster to the DNS servers or IP addresses in their mapping algorithm ● The content owned CDNs also include signals from the content clients in the algorithm Latency and other QoS data ● The load of the individual servers or clusters is part of the decision of exactly which server the end- user will end up using. ● Load balancing can take place in the cluster or in the CDN response to (where is?) Load ● Not all content is on all Embedded servers ● Content that is pushed out is distributed according to CDN magic ● Not all content from a CDN will be cached when the content is pulled ● This means where the content is/can be is also part of the steering decision. Content
  • 14. Traffic profiles for different ways of placing content • Centrally calculated placement • Fill window • Optimized to reduce edge fill traffic as much as possible Typical Netflix traffic profile in well dimensioned deployment
  • 15. Traffic profiles for different ways of placing content • Proxy and akamai magic • Relatively dynamic content • Not all content will be at the embedded servers Typical Akamai traffic profile in well dimensioned deployment
  • 16. Traffic profiles for different ways of placing content • Proxy and Facebook magic • Relatively dynamic content • Only the static is cached • Not all content will ever be at the embedded servers Typical Facebook traffic profile in well dimensioned deployment Feeds Likes Anything that needs to be globally consistent
  • 17. Limit who can use the Servers Most embedded Servers will have access list that limits who will be served by the cache The common source to generate these is a BGP feed from the host ISP Some will use one single feed per ISP, other per cluster or per cache. There must be consistency between the user- group that can be mapped to a cluster or server and the user-group that are allowed to use the cluster or server. This means there is a requirement for the CDNs where each server has a BGP sessions that the announcements to each server in a cluster is identical.
  • 18. Deploying embedded servers deep into the network • Understand if the end-user mapping is supported by your network implementation • DNS servers • IP address plan • Use of CGN • Be aware of clustering • Off load/hit rate for the local cluster might be lower than for a big cluster because of less disk space in the cluster
  • 19. And what about the hardware? 1-2U 10G /50G/100G Interfaces Choice of AC and DC power It varies whether a CDN managed switch is required optional or not applicable. Installation process are more/or less automated, but some still require the ISP tech to do basic configuration of the CDN equipment. Evolves continuously: less space, more throughput, more and faster storage Interfaces supports the industry standard at any given time
  • 20. Is it easier or harder in 2023 compared to 2012? ● Large boxes ● Off load unpredictable ● Few players ● No CGN support 2012 ● Small boxes ● More predictable offload ● More predictable and flexible end-user mapping ● Many players ● Some CGN support 2023 CONCLUSION
  • 21. What do you experience?
  • 22. THANK YOU Jack Fang Solutions Engineer - APAC jfang@kentik.com